EP0946820B1 - Method and device for conducting a process in the production of paper - Google Patents

Abstract

In order to produce paper from fibrous material, a prepared substance is conveyed to a paper machine (15) with a head box. The substance can contain, for example, a mixture of different recycling paper types with cellulose. In accordance with the invention, at at least one point on the prepared substance and/or the paper (200), continuous spectra of electromagnetic radiation and/or continuous spectra of mechanical properties are measured on the substance (A, B, C, D) and/or paper (E, F). Characteristic quantities are derived from these spectra and fed into a state model, from whence a process model is, in addition, derived with the process qualities. In certain cases, discrete physical and/or chemical properties are also used to form the model. The state model and/or process model are used for conducting and optimising the process.

Description

The invention relates to a process control method and process optimization in the manufacture of paper made of fibrous material, in particular waste paper and / or cellulose, by processing the pulp and then forming it of the paper using at least one state model and / or process model. The invention also relates to an apparatus for performing this method.

Nowadays, paper is used as a raw material for the production of paper Portions of waste paper and separately produced cellulose used. The waste paper is usually dissolved and Deinken, the pulp by removing the residual lignin processed, the mixed product generally referred to as "substance" becomes. For a paper making plant, including stock preparation and paper machine, are involved in litigation, especially litigation with automation devices, significant requirements.

The process control in papermaking is therefore difficult because of mechanical paper properties - such as Tensile strength, tear propagation resistance, flat crush resistance, Ring crush resistance, the burst strength, pick resistance, Behavior towards liquid, printability - , can only be recorded on the finished product. This one Values measured in the laboratory with a time delay can be a interim incorrect production cannot be excluded. In order to avoid wrong productions as far as possible, usually higher quality raw materials and auxiliary materials, e.g. glue, Additives used on a larger scale than actually would be necessary to achieve the desired paper quality.

• Der Prozeß wird nach Erfahrungswerten gesteuert. Regeleingriffe sind nur in geringem Umfang möglich. Die Produktqualität kann nur im nachhinein bestimmt werden. Eine große Streuung der Qualitätsparameter ist daher häufig nicht vermeidbar, insbesondere wenn die Qualität der eingesetzten Rohstoffe wie die des Altpapiers oder die des Zellstoffes stark schwankt.
• Es wird bereits versucht, den Prozeß über online meßbare Hilfsgrößen zu regeln. Diese Strategie wird nur gelegentlich eingesetzt, da die Zahl der mit nicht zu hohen Aufwand meßbaren Hilfsgrößen begrenzt ist und zudem häufig nur ein empirischer Zusammenhang mit den Zielgrößen besteht.

The following strategies are often used in the prior art:

• The process is controlled based on experience. Regular interventions are only possible to a limited extent. The product quality can only be determined after the fact. A wide spread of the quality parameters is therefore often unavoidable, especially if the quality of the raw materials used, such as that of the waste paper or that of the pulp, fluctuates greatly.
• Attempts are already being made to regulate the process via auxiliary variables which can be measured online. This strategy is only used occasionally because the number of auxiliary variables that cannot be measured with too much effort is limited and, moreover, there is often only an empirical connection with the target variables.

The following publications are known about the use of infrared spectroscopy in papermaking:
In DE 195 10 009 A1, selected spectral characteristic values are used for a process for the process control of a paper machine for the determination of product quality parameters with neural networks. Using the example of the headbox of a paper machine, neural networks are used that have been trained on the product parameters measured in the laboratory. DE 195 10 008 A1 describes a process control process in pulp and paper production, which uses spectral parameters of different wavelengths to determine the starting materials and to evaluate the quality of the raw materials. Correction signals for the regulating or control device are derived therefrom, in particular with neural networks. The spectral parameters are measured on the raw material "wood", in particular wood chips, wood fibers, or on the raw material "waste paper" as raw materials for pulp and paper production.

WO 95/31709 A1 makes a prediction of the wet tear resistance with the help of spectroscopic measurements on paper or pulp using mathematical methods. WO 93/15389 A1 describes a method and a Device for inline determination of the so-called "freeness'", i.e. the degree of refining of refiner wood pulp is known: spectral characteristics of selected wavelength bands processed by a calculation process around the desired quality index Get "Freeness". When calibrating at the same time spectral measurements and quality measurements The measurement results are processed in new variables and using the dependency of "freeness" the main component method is determined using the Wavelengths are between 0.1 - 10 µm.

WO 95/08019 A1 deals with the processing of waste paper, i.e. to the production of so-called deinked pulp, where spectral and / or physical on the waste paper suspension Characteristic values are recorded. As spectral parameters become more spectral after irradiation with light Distribution of special intensities at defined wavelengths detected and evaluated in particular via a neural network.

Finally, DE 41 33 439 A1 describes a method for continuous measurement of the so-called dynamic water retention a coating on a passing support and especially described on a piece of paper in which radiation at a first station with a known frequency spectrum in any level in terms of direction the expiry of the carrier is irradiated and in the same Plane the reflected radiation is detected. By specific Procedure of the measurement, in particular repetitive measurement, is from the reflected signals Retention determined according to a specific rule.

Based on the state of the art, it is a task the invention, the process control in papermaking further optimize and an associated device too create.

The object of the invention is with the features of the claim 1 solved, the dependent claims being advantageous Include further training. The associated device is specified in the only property claim.

Spectroscopic measurements are preferably used in the invention online in the spectral range from 0.1 µm to 400 µm Fabric and / or on paper, i.e. specifically on a stock suspension in the stock preparation and / or on the headbox and / or carried out on the running paper web. The spectra can be evaluated using different calculation methods and allow a prediction of the achievable Paper quality, the influence of auxiliaries, e.g. Glue, also taken into account in the prediction models can be. Based on this prediction, the Process can be intervened regulatingly, e.g. B. by change the raw material mixture, reducing the proportion of high quality Raw materials, change in grinding or the addition of Auxiliaries, e.g. of glue. This can advantageously a saving in the use of raw materials with the same Product quality and avoiding wrong productions inferior quality can be achieved.

In addition to predicting properties that were previously only measurable offline properties such as ash content, whiteness, Opacity, dirt spots directly determined and / or predicted be the z. Currently online only with a special, in part very complex sensors can be measured. An essential one The advantage of the invention lies in the fact that it is saved additional sensors.

The invention is therefore compared to what was initially discussed State of the art in all essential points, such as the Location, method of measurement, processing of spectra, and Advantageously developed the modeling.

Within the scope of the invention, the measurement can also be carried out on a With the help of a sampler. The The sample can be processed in the desired way, e.g. by concentrating, drying and / or forming a Sample sheet. When forming a sample sheet is one subsequent laboratory measurement of the paper properties possible, e.g. to control the quality of the predictions and, if necessary, the models for quality parameters improve if e.g. important process variables such as raw material quality change or the process flow changed has been. Alternatively or in parallel, a measurement on the running paper web, e.g. before the dryer section or at the end of the paper machine in front of the so-called drum of the associated one Winder.

As already mentioned, the measurement is carried out in particular with electromagnetic waves in the wavelength range of 0.1 µm - 400 µm. The absorption, Emission, luminescence or Raman spectra of the material samples. Absorption spectroscopy can be diffuse in transmission Reflection or attenuated total reflection (ATR = "attenuated total reflection "). The excitation of the luminescence can e.g. through the radiation of electromagnetic Radiation, e.g. UV radiation, or by a specific chemical reaction such as chemiluminescence or the like, the In contrast, excitation of the emission can e.g. by irradiation with Electrons take place.

When measured in the range of I nfra r soldering (IR: 1 .mu.m to 25 .mu.m) may preferably F ourier- T ransformations- spectroscopy (FTIR) can be used. On inhomogeneous material samples, the spectroscopic measurement can be carried out several times or different measuring locations on the dissolved or prepared fiber material and / or on the paper.

• durch Fourier-Transformation
• bei Messung der Absorption durch diffuse Reflexion Umrechnung in Kubelka-Munk-Einheiten und Korrektur von Mehrfachstreueffekten
• durch Normierung und Glättung der Spektren
• durch Ermittlung von für die Modellbildung ungeeigneten Spektren. Bei der Modellierung der z.B. mechanischen Papiereigenschaften sind Spektren, die z.B. an Schutzpunkten, z.B. Farbreste im Altpapier usw., gemessen wurden, ungeeignet und müssen verworfen werden. Die Ausschaltung ungeeigneter Messungen kann z.B. durch Vergleich mit Referenzspektren erfolgen.
• Bei mehreren Spektren zu einer Probennahme durch Bildung von Mittelwerten.

The spectra can advantageously be preprocessed by the following measures:

• through Fourier transform
• when measuring the absorption by diffuse reflection, conversion into Kubelka-Munk units and correction of multiple scattering effects
• by normalizing and smoothing the spectra
• by determining spectra that are unsuitable for modeling. When modeling the mechanical paper properties, for example, spectra that have been measured, for example, at protective points, for example color residues in the waste paper, etc., are unsuitable and must be rejected. Unsuitable measurements can be switched off, for example, by comparison with reference spectra.
• For several spectra, take a sample by averaging.

• Hauptkomponentenanalyse (PCA), bei dem eine Beschreibung der Spektren durch die Hauptkomponenten erfolgt
• "Partial Least Square"(PLS)-Verfahren
• Neuronale Netze
• Analytische Beschreibung der Spektren, z.B. im Bereich des IR durch Lage, Intensität und Breite der wichtigsten Absorptions- oder Emissionspeaks, Ermittlung dieser Größen z.B. mit einfachen Minimum-Maximum-Verfahren oder der 2. Ableitung.

Die Kenngrößen werden zur Modellierung der gewünschten Qualitätsparameter herangezogen.After these first processing steps, the following computational methods for determining characteristic values can preferably be used for further processing of the spectra, which are prepared in whole or in sections:

• Main component analysis (PCA), in which the spectra are described by the main components
• Partial Least Square (PLS) process
• Neural Networks
• Analytical description of the spectra, for example in the area of the IR by position, intensity and width of the most important absorption or emission peaks, determination of these quantities, for example using a simple minimum-maximum method or the second derivative.

The parameters are used to model the desired quality parameters.

The models for calculating the quality parameters can be found at sufficiently large number of data, preferably neurons Networks, fuzzy systems or multilinear regression models or combinations thereof. As an alternative to purely data-driven ones Combined models are also possible, where analytical knowledge is introduced, e.g. that with increasing ash content of the raw material the mechanical quality parameters the paper produced from it deteriorate. To increase the accuracy of the forecast, it makes sense to e.g. to form five to ten sub-models, their assignment e.g. about the raw materials used.

The list of models, i.e. Selection of model input variables, and their validation takes place with laboratory measurements on Intermediate and final product. Neural networks can be used be trained online. One in the Spectra preprocessing integrated method of the so-called "Novelty Detection" according to the older, unpublished DE 196 322 45 A1 can in the running process in time the need for a new training phase Show.

Figur 1

ein Schema der Meßstellen bei einer Anlage zur Papierherstellung,

Figur 2a bis 2c

kontinuierliche Spektren von optischen Messungen an einer Faserstoffsuspension und von mechanischen Messungen an Fasern,

Figur 3

ein Zustandsmodell für die Qualitätsparameter bei einer Papiermaschine,

Figur 4

ein Prozeßmodell des sogenannten CMT-Wertes einer Fasermischung bei der Altpapieraufbereitung,

Figur 5

ein Prozeßmodell des CMT-Wertes des Papiers auf der Papiermaschine,

Figur 6

ein dynamisches Prozeßmodell für den CMT-Wert,

Figur 7

den schematischen Aufbau einer Prozeßoptimierung zum Steuern der Papierherstellung,

Figur 8

eine Variante von Figur 7 unter Einbeziehung des dynamischen Modells gemäß Figur 6,

Figur 9

die Vorverarbeitung und Verdichtung der Spektren und

Figur 10

eine Vorrichtung zur optimierten Prozeßführung bei der Papierherstellung.

Further details and advantages of the invention result from the following description of the figures of exemplary embodiments with reference to the drawing. Show it

Figure 1

a diagram of the measuring points in a plant for paper production,

Figure 2a to 2c

continuous spectra of optical measurements on a fiber suspension and of mechanical measurements on fibers,

Figure 3

a state model for the quality parameters in a paper machine,

Figure 4

a process model of the so-called CMT value of a fiber mixture in waste paper processing,

Figure 5

a process model of the CMT value of the paper on the paper machine,

Figure 6

a dynamic process model for the CMT value,

Figure 7

the schematic structure of a process optimization for controlling the paper production,

Figure 8

7 shows a variant of FIG. 7 including the dynamic model according to FIG. 6,

Figure 9

the preprocessing and compression of the spectra and

Figure 10

a device for optimized process control in papermaking.

The figures are partially described below together. The same or equivalent parts have corresponding Reference numerals.

1 shows an example of a measuring point diagram for the Paper production shown, in which the production of the Paper is made using waste paper. For that is in 1 shows a mixing chest with 10, for example three types of waste paper via a first pulper 1, one second pulper 2 and a third pulper 3 are supplied. Pulpers 1 to 3 are followed by stock preparation units 11 to 13. From the mixing chest 10 the processed one Substance on the headbox not shown in detail given a paper machine 15, in the known Way the paper is formed. The paper machine 15 gives a continuous paper web 200 as a production product out.

FIG. 1 shows the measuring point diagram of the waste paper preparation and the paper machine in detail. Three measuring points for spectra A, B, C are introduced into the mass flow m ˙ ₁ , m Meß ₂ and m ˙ _{3 of} the processed substance. After the mixing chest, a spectrum D is taken from the current m ˙ _M. Further spectra E and F can be recorded on paper machine 15 on the one hand during paper formation and on the running paper web 200 on the other hand.

For process control, the measured values in the material flow are from Importance. This can be done according to the sorting and cleaning stages before the headbox a not shown in detail Samplers can be installed. In this sampler formed a sample sheet for paper formation, which in particular can also be used for laboratory measurements. In the sampler an FTIR spectrometer is integrated with which under Exclusion of air humidity, e.g. by nitrogen flushing in a vacuum, in the IR range (1 to 25 µm) the sample sheet several times, for example 10 to 20 spectra, on different Spectroscopy measuring sites in transmission or diffuse reflection can be.

In FIG. 2a, 21 indicates the course of an IR spectrum of paper. A section according to FIG. 2b in the range from 3700 to 3600 cm ⁻¹ (so-called ash range) of the spectrum is particularly important for practice. In contrast, FIG. 2c shows the spectrum of a fiber length distribution of fiber or paper, as can be seen from an example measurement of the mechanical properties. There is approximately a course along the boundary line 22 of the bar representation corresponding to a so-called Bernoulli distribution. The distribution of the sieve fractions of individual fibers of the fibrous material has a similar course.

The spectra obtained using FIGS. 2a and 2b are prepared using a wide variety of mathematical methods. This essentially involves preprocessing of the spectra to bring analytical knowledge to a possible outlier detection and the intended Deriving control variables for the operation of the entire process engineering Investment.

The measured spectra are first smoothed and standardized. The optical spectra according to FIG. 2a are advantageously divided into three sub-spectra, the "OH range" (3800-2600 cm ^-1 ), the "fingerprint area" (2600-600 cm ^-1 ) and the "ash range" (3700-3600 cm ^-1 ). The "OH area" and "fingerprint area" are each standardized individually, namely the "OH area" to the minimum at, for example, 3750 cm ^-1 and the maximum at 3420 cm ^-1 , the fingerprint area to the maximum at 2900 cm ^-1 and the minimum at 1920 cm ^-1 .

The "ash area" between 3700 and 3600 cm ^-1 is subjected to a baseline correction for the following processing. The absorption maxima are determined as peaks in the spectra using a minimum-maximum method and the second derivative, peak position, peak intensities, peak width and other values being recorded. Spectra identified as useful are used for averaging.

The ash content is calculated from the "ash area". This is considered one of the direct quality parameters of the Product saved. On the other hand, the value is considered one of the input variables for the later modeling.

The calculated mean spectra are then separated according to "OH area" and "fingerprint area" and broken down into its main components. With e.g. a multilinear regression using selected main components as input variables the achievable Quality parameters, e.g. the burst strength, the CMT or predicted the tearing length. For everyone to be calculated Quality parameters are to be created in own models. to Increasing the prediction accuracy will be for each parameter Sub-models formed. The assignment to the sub-models takes place based on the assignment of the measurements to substance classes during preprocessing. When assigning Up to three substance classes for measurements on one sheet corresponding combination models are calculated, e.g. a linear mix of models according to the proportion of Class of substance in the sheet. The predicted values are as To mark the result of a combination model.

Based on this information, it is possible to determine the raw material composition to optimize or use To regulate auxiliary materials, such as glue in particular.

In FIG. 3, 30 denotes a state model of the paper machine, from which the quality parameters for paper such as fiber length, tear length, burst pressure, tear resistance and the like are essentially derived. For this purpose, from spectra according to figure 2a and / or figure detected with either 2b of the so-called principal component analysis (PCA = p rincipal c omponent a nalysis) or with the so-called PLS method (p artial l east s quare) parameters, and input to the model. The parameters PC1 to PCn are shown, for example, which were determined with, for example, 10 scores. Furthermore, discrete mechanical or chemical properties are also selected and entered into the model 30. With an additional input of the process properties, the actual state model becomes a process model.

To make the modeling more flexible and to make the prediction more precise to increase, the ones already mentioned Classes of substances are formed based on the composition of the substance. The substance classes can, for example, according to the raw materials, by type of waste paper or type of fiber such as TMP, cellulose or sanded wood, or according to the products, e.g. classified according to the basis weight or the type of paper produced, be divided.

Ideally, each measurement should only be for a single sheet can be assigned to a substance class. Can do more are classified as three substance classes Alarm, because the mixture of substances is obviously very inhomogeneous is.

The fabric differs only slightly from the known compositions ab, but is not yet known to the system, it will Spectrum used for evaluation. The system does, however a message that a new measurement of the quality parameters in the laboratory for a new learning phase.

If the measured spectrum deviates significantly from the references, is the measurement e.g. at a dirt point. The spectrum is then not used for further calculations used, but saved for documentation purposes. It turns out that too much of the measurements e.g. 25% must be discarded, then the system returns Alarm, e.g. an error in the previous sorting and Cleaning levels can be present.

In Figure 4 is a process model 40 specifically for waste paper processing as a sub-process in papermaking shown. The following are entered here for the individual waste paper preparation lines the flow, the optical Spectrum and fiber length distribution, being specifically according to Figure 3 is proceeding. The result is the so-called CMT value of the waste paper mixture.

Similarly, in Figure 5 is a process model 50 of the paper machine shown in which the input variables, the flow, the optical spectrum, the mechanical spectrum of the Fiber length distribution, the amount of glue and the temperature of the Glue can be chosen. Here is the output the CMT value of the paper, the burst pressure, the tear length and other significant sizes.

In Figure 6 it is shown that with a dynamic specification of the input variables at time k, ... k-n and additionally previously known CMT values at time k, ... k-n a dynamic process model 60 of the paper machine itself Time-variable values such as CMT value, burst pressure and tear length for paper at time k + 1.

When setting up the models can be of neural networks and / or fuzzy logic methods can be used. In particular, it is about the validity of the models validate what is done through an online individual training Models or the partial models can be made. It can be for practice will be important through computer-aided selection of all information-carrying data a check of each results obtained. This procedure was called So-called "Novelty Detection" proposed and enables new Bring data records into the evaluation process. If available Results that are not consistent are retraining of the models necessary.

The sizes thus obtained are used for process control. For this purpose, the process is generally designated 70 in FIG. 7, which gives the current process status based on the spectra with 71 results. The process model is designated 72 here from which the data into a unit for cost function 73 be given at the same time as data for costs and Prices from unit 74 is applied. An optimizer From this, 75 determines the manipulated variables 76 that are used in the process model 72 are fed back and continue to be the optimal ones Actuating variables 77 for process control. This can be done via a Switch 78 can be entered by the operator, provided that recognized as meaningful.

The corresponding results from FIG. 8 for process control, at which equally a dynamic model accordingly 6 is used. Here is additional a unit 79 with the dynamic model in which the current process status on the one hand and the optimal ones On the other hand, manipulated variables can be entered.

FIG. 9 shows that a unit 91 for preprocessing and compression of the entire spectrum from which the parameters are calculated in accordance with unit 92. The Parameters flow into state model 93 and into the process model 94, with additional discrete mechanical and / or chemical properties and the process status description complement the state model to the process model. From the state model 93 becomes the quality parameter for the fiber and from the process model 94 the quality parameters for derived the final product.

Figure 10 shows how using a digital computer 105 with appropriate evaluation and optimization software in the entire process control system is integrated. In doing so generates optimized manipulated variables that a known automation device 100 act as a process control system, which in familiar with the actual plant for implementation the process interacts. Basically are the usual systems with several spectrometers 101 to 103 and an associated software package that on usual computers runs, added.

Claims (27)

1. Method for process management and for process optimisation in the production of paper from fibre pulp, in particular from recycled paper and/or chemical pulp, by treating the fibre pulp and subsequently forming the paper in a paper-making machine while using at least one state model and/or process model, having the following features:

   a) at at least one point, continuous spectra of electromagnetic radiation and/or continuous spectra of mechanical properties are measured from the treated fibre pulp and/or from the paper,

   b) characteristic quantities (PC1, ..., PCn) for the treated fibre pulp and/or the paper are determined by mathematical evaluation of the continuous spectra,

   c) the characteristic quantities (PC1, ..., PCn) and the laboratory measurements of the product properties are entered into the state model and/or the process properties are additionally entered into the process model, by means of which the models are verified,

   d) optimised manipulated variables for an existing process control system are formed with the aid of the state model and/or process model set up in this way.
2. Method according to Claim 1, characterised in that at at least one point, discrete physical and/or chemical properties are recorded from the fibre pulp and/or from the paper.
3. Method according to Claim 1 and Claim 2, characterised in that the discrete physical and/or chemical properties are additionally used for setting up the state model and optionally the process model.
4. Method according to Claim 1, characterised in that measurements are taken at wavelengths of the electromagnetic radiation of between 100 nm and 400 µm.
5. Method according to Claim 4, characterised in that the electromagnetic radiation is recorded in an absorption, emission or luminescence spectrum or as a Raman spectrum.
6. Method according to Claim 4, characterised in that the electromagnetic radiation is recorded in transmission, direct or diffuse reflection or attenuated total reflection (ATR).
7. Method according to Claim 1, characterised in that the fibre length distribution or the screen fractions of the fibres are employed as spectra of the mechanical properties.
8. Method according to Claim 1, characterised in that recycled paper from at least one recycled-paper line is used for the production of paper, the number of recycled-paper lines that are connected in parallel being unlimited.
9. Method according to Claim 1, characterised in that the measurement of the continuous spectra is carried out at the output of the recycled-paper treatment and/or after the pulp-treatment and/or in the region of the paper-formation in the paper-making machine.
10. Method according to Claim 1, characterised in that a principal component analysis is carried out on a predetermined number of spectra and, for data reduction, a corresponding number of scores is selected, and in that the characteristic quantities for the model formation are determined therefrom.
11. Method according to Claim 9, characterised in that the spectra are pre-processed and compressed, and in that the specific characteristic values of the spectra, in particular the principal components, are selected for describing the product state and are entered directly into the state model.
12. Method according to Claim 9, characterised in that the spectra are pre-processed and compressed, in that their specific characteristic quantities, in particular the principal components, are introduced into the state model, and in that the product properties are formed at the output of the state model and are entered directly into the process model.
13. Method according to Claim 1, characterised in that spectra that are unsuitable for the model formation are eliminated by plausibility testing.
14. Method according to Claim 1, characterised in that the consistency, the degree of freeness, the flow rate of the fibre suspension and/or the basis weight and the moisture of the paper are recorded as discrete physical and/or chemical properties.
15. Method according to Claim 1, characterised in that the characteristic quantities that are obtained by evaluating the continuous spectra are employed for controlling and/or regulating the paper-formation process.
16. Method according to Claim 14, characterised in that the characteristic quantities that are obtained by evaluating the continuous spectra are employed for controlling and/or regulating the treatment stages, the treatment stages being optimised individually or even in combination.
17. Method according to Claim 1, characterised in that the quality parameters of the treated pulp, such as in particular the whiteness, dirt specks, tearing length, are modelled in order to control and/or regulate the treatment stages.
18. Method according to Claim 16, characterised in that the model approaches are also employed for calculating the use of the auxiliary materials, such as glue or chemical utilization, besides predicting the product quality.
19. Method according to Claim 16, wherein control and/or regulation of the paper production is carried out, characterised in that the model approach is used with the quality parameters in a process optimisation.
20. Method according to one of the preceding claims, characterised in that a cost function is formed that is optimised by an optimiser through suitable variation of the manipulated variables.
21. Method according to one of the preceding claims, characterised in that the optimisation is carried out using genetic algorithms.
22. Method according to Claim 19, characterised in that a cost function for the production costs and/or a profit function is used as the cost function.
23. Method according to one of the preceding claims, characterised in that as a dynamic model is used for checking the manipulated variables that are optimised by a static model, a neural network being used as the dynamic model which is used.
24. Method according to one of the preceding claims, characterised in that the model and/or the submodels are trained online.
25. Method according to one of the preceding claims, characterised in that a check of the results that are obtained ("novelty detection") is carried out by computer-assisted selection of all data carrying information.
26. Method according to Claim 25, characterised in that retraining is carried out when there are inconsistent results.
27. Device for carrying out the method according to one of Claims 1 to 26, consisting of at least one spectrometer (101, 102, 103) for measuring continuous spectra, a digital computer (105) for mathematical evaluation of the continuous spectra with a view to determining the characteristic quantities (PC1 ... PCn) and for setting up the state model and/or process model from the characteristic quantities (PC1 ... PCn) and optionally the process properties, as well as a process control system (100) for optimising the process for the production of paper from fibre pulp with the use of optimised manipulated variables.

Images